ESTIMATION OF NON-OPTICALLY ACTIVE WATER QUALITY PARAMETERS FROM SENTINEL-2/MSI AND LANDSAT-8/OLI SENSORS

Abstract

Monitoring water quality parameters in reservoirs is required by law in Brazil, especially considering the increasing demand for water resources and their multiple use. Passive remote sensors measure the reflectance of solar radiation on the water surface and can allow non-optically active parameters to be evaluated through their relation with optically active parameters. Considering that empirical models are rarely used in areas other than those for which they were conceived, the objective of this paper is to apply previously developed non-optically active parameter models (October/2019) to image data from a later date (May/2020) for the same reservoir (Três Marias Reservoir/MG) and, hence, assess the replicability of these models. The approach uses the same spectral bands as predictors in the multiple regression models developed by Pizani et al. (2020) to estimate three parameters: electrical conductivity, total dissolved solids and dissolved oxygen. The results indicate high coefficient of determination between the data measured in situ and the spectral reflectance, highlighting the better performance of the MSI/Sentinel-2 sensor (r² > 0.71) compared to the OLI/Landsat sensor (r² > 0.50). These results show advancement in remote sensing research to monitor water bodies while reducing surveying costs.